Active Landmark Navigation Using a Pan-Tilt Sensor System for Unmanned Aircraft

Abstract

This paper presents an approach on how to integrate pan-tilt camera systems into the state estimation of unmanned aircraft to introduce an additional source of navigation data. The proposed state estimation algorithm based on an Unscented Kalman Filter is presented with an additional augmentation of the measurement model to cope for the uncertainties introduced by the pan-tilt unit of the camera system. Furthermore, the correlation of the visual measurements due to the common pan and tilt axes of the camera system is addressed. Second, the advantages of using an actively controlled pan-tilt camera system in the context of landmark-based navigation are shown. The navigation performance of such an actively controlled camera system is compared against fixed camera systems. Furthermore, the focal length aka. zoom is taken into account to further motivate the use of those actively controlled pan-tilt camera systems. The comparison is based on a Matlab/Simulink simulation environment providing measurements like inertial, magnetic or barometric sensor data as well as emulated landmark tracking data. Finally, an outlook on future flight tests using an 85 kg unmanned helicopter is given.